The KnowEMS initiative hosted a webinar examining the use of artificial intelligence (AI) in emergency care and prehospital settings, bringing together perspectives from international organizations and frontline EMS providers. The session featured presentations by Oleg Storozhenko (WHO Europe), Daniel Muñoz (SAMUR-PC Madrid), and Chaim Rafalowski (Magen David Adom, Israel; ESCORT project).
Speakers noted a growing wave of initiatives integrating AI into EMS workflows—from resource optimization and coordination to information sharing and operational support. WHO Europe highlighted digitalization of Emergency Medical Teams (EMTs) as a strategic priority within its EMT plan, including piloting comprehensive tools such as KIMEP that span from preparedness (e.g., roster management) to real-time operational support. In some deployments, use of these tools during mentoring and verifications is mandated, including coordination with the European Commission for teams engaged in RescEU activities.
A recurring theme was the high variability across countries and organizations: some adopt off-the-shelf solutions, while others build bespoke systems from scratch. This heterogeneity—paired with the rapid evolution of AI technologies—makes a harmonized approach difficult at present. Participants stressed the need for interoperable standards without stifling innovation, especially given different norms, legal frameworks, and operational contexts across Europe and beyond.
Most tools today concentrate on the “productivity” side of EMS operations: optimizing resource use, improving coordination, and streamlining information flows. WHO’s KIMEP was presented as a comprehensive example currently under pilot, illustrating how platforms can support preparedness and live operations in complex, multi-agency environments.
Panelists underscored an unavoidable rule: AI is only as good as the data it receives. Ensuring high-quality inputs requires both robust IT tooling and human oversight. Poor data integrity leads to unreliable recommendations—“garbage in, garbage out”—posing risks in time-critical clinical contexts. Clear data definitions, purpose limitations, and use constraints were emphasized; in current EMS pilots, patient-level data is generally not collected, and clinical data is often aggregated to minimize privacy and compliance risks.
The discussion highlighted pressing legal updates needed to align with AI realities, including:
Achieving effective implementation requires multi-stakeholder consensus, which is challenging in cross-border or multi-institutional deployments.
Participants called for ethical deliberations to proceed in parallel with technical development. Clinical care extends beyond vitals and test results; socioeconomic factors, support systems, health literacy, and caregiver capacity all shape patient vulnerability and outcomes. Future AI tools should contextualize recommendations within these realities, acknowledging that ethical and societal considerations are context-dependent and cannot be universally standardized without nuance.
Emergency care often involves multiple concurrent diagnoses (e.g., fracture plus concussion) layered on pre-existing conditions (age, medications, chronic disease). The clinical eye—expert human judgment—remains essential. Present AI tools do not yet handle multi-morbidity holistically; they are typically focused on narrow, specific diagnostic or operational tasks.
Projects such as NIGHTINGALE have pointed out a critical barrier: no comprehensive database exists to enable robust machine learning across the acute care journey, from prehospital to intra-hospital phases. Although patients may be continuously monitored, those streams are often not recorded cohesively. The ESCORT project is addressing a related gap: even when vitals are recorded, they’re not consistently coupled to medical history, physical examinations, and treatments—making it hard to attribute changes in vitals to specific interventions or clinical context.
The webinar concluded that AI can enhance EMS performance, but safe, effective deployment depends on high-quality data, strong governance, ethical safeguards, and a clear understanding of AI’s current limitations. Continued collaboration across countries and disciplines will be essential to build systems that support clinicians, respect patients, and improve outcomes in the most time-critical care settings.